An AI Halloween Market Scare
(…) Google’s advertising revenue during the latest quarter undershot investor expectations, raising worries that AI may be cannibalizing its search engine. Search ads are a key source of Google revenue, which it has used to fund its AI build-out. If more people use AI large-language models for search, that could reduce Google’s income and make it harder and costlier to fund its AI ambitions.
There are also lingering concerns that Google may struggle to monetize its own large-language model Gemini and AI search tools to the degree it has traditional search. But on the bright side, revenue from Google’s cloud unit doubled from a year earlier from increased demand for its AI tools and custom chips, which could pose a competitive threat to Nvidia and its high-cost chips.
Intel had an even worse day than Google, with the chip-maker’s stock price plunging 29% on Friday. AI was indirectly to blame. Surging demand from hyperscalers like Google and OpenAI has driven up prices for memory chips and, consequently, devices like PCs and smartphones. This is hurting the market for PCs, which run on Intel’s legacy chips.
Amazon’s stock on Friday also took a header after the company reported higher than expected capital spending on AI and negative cash flow. This amplified market concerns that the AI build-out is proving more expensive than anticipated.(…)
At some point hyperscalers may have to pull back, which could hurt the entire AI ecosystem. That might hit chip makers like Nvidia and Micron Technology that are currently churning out huge profits.
This in large part explains the routs in other AI-related stocks, which dragged the Nasdaq down 3% on Friday. A correction was probably inevitable given the waves of capital that have been pouring into any company claiming to have an AI angle. (…)
Interestingly, the “AI correction” came after AI players reported weaker than expected growth in their respective legacy businesses: search for Google, retail sales for Amazon and end-user hardware for Intel.
Their respective AI businesses kept booming along, requiring investments that investors worried might not be sufficiently financed by legacy cashflows.
Lenders and investors are getting choosier about where to place their bets and are demanding tougher terms, signaling a new phase in the AI build-out. Among their concerns: the prospect of delays in big projects as a grassroots political backlash to data centers gains traction, and the threat that AI-related profits from hyperscalers and startups such as Anthropic and OpenAI won’t materialize as fast as promised.
Plus, a flood of stock and bond offerings is compounding concerns that investors are getting overexposed to AI.
The new terms of trade are visible in every corner of markets. Bonds issued even by big tech companies such as SpaceX and Meta are trading at steep discounts. AI cloud companies such as CoreWeave, Nebius and Nscale that borrow big to build and sell capacity have effectively been shut out of the bond market in recent weeks, people familiar with the matter said.
Taken together, it all amounts to a buyer’s market for Wall Street. Even powerful players such as Broadcom are having to sweeten the terms offered to banks to secure loans. (…)
The stock market is starting to reflect investors’ nervousness that the AI payoff might not arrive—at least in the near term. (…)
There is still money available for creditworthy companies. SpaceX this week announced it raised the biggest-ever financing for AI chips, a $40 billion package that will allow it to purchase the latest lineup of Nvidia chips.
Its record only lasted a few hours, when chip-design giant Broadcom sealed a $63 billion deal to lease chips to Anthropic. (…)
Meta, Amazon, Alphabet and Microsoft carry top credit ratings and have significant room to pile debt on their balance sheets. But after many of them flooded the high-grade debt market with new bond sales this year—and signaled they plan to keep doing so for the foreseeable future—the investment firms, insurance companies and pension funds that lend to them are feeling increasingly tapped out, and concerned about their concentration in one sector. (…)
Add the odds of a Democrat sweep Tuesday and a AI-hot economy likely to push the FOMC to hike rates again in December and maybe after and there is enough uncertainty to make investors reevaluate their positions.
And the US/Israel war with Iran and its impact on supply chains, commodities and inflation.
And tariff/immigration policies pushing corporate costs up.
And the US government deficit needed financing.
In fact, one might say that the US government has set the stage for the current reckoning that it ain’t good to be small if you must
- switch suppliers and reorganize your supply chain,
- absorb tariff and energy costs,
- hire skilled people,
- build AI capabilities,
- secure limited inputs and
- find affordable financing.
Is OpenAI big enough?
OpenAI has raised less than half of the $30 billion it had initially targeted for its latest funding round, with market jitters over the AI trade dampening investor appetite for one of the industry’s biggest names.
After receiving tepid interest from some large investors, OpenAI has so far secured roughly $11 billion in commitments for the new financing, according to people familiar with the matter. The deal values the company at $1.25 trillion—less than the $1.4 trillion valuation that it had earlier hoped to achieve. (…)
The maker of ChatGPT and Codex is now hoping to raise up to $20 billion, one of the people said. (…)
AI Adoption Rises to 24.6%
From Goldman Sachs:
AI adoption by firms rose by 0.8pp to 24.6% among US establishments (with adoption expected to rise to 28.6% in the next 6 months) according to the Census Bureau’s Business Trends and Outlook Survey (BTOS).
Adoption is approaching 50% in the information, professional, and finance sectors. At a detailed subsector level, computing, securities, and publishing firms continue to lead adoption.
Over half of establishments with over 250 employees expect to be using AI in the next six months.
Other business surveys note that firms in Europe and Asia reported the highest rates of positive return from their AI initiatives.
Amid the debate on AI adoption, ROI and productivity, one can see the very rapid positive trends from these surveys:
- A McKinsey August survey said that 40% of large organizations have adopted AI agents and are scaling their deployment throughout the organization. 80% of respondents said AI has improved their individual productivity.
- A KPMG September survey found that 62% of organizations are now building, deploying or developing AI agents, up from 53% last quarter. General workforce adoption of AI tools rose to 44% from 23% last quarter.
- A Dun & Bradstreet October survey revealed that 28% of surveyed businesses report broad or strong returns from AI efforts. Data readiness is strongly associated with AI ROI. Only 8% reported that their enterprise data is ‘fully ready” to support AI at scale, 44% as “partially ready” and 33% as “mostly ready”. 65% plan to increase AI spending over the next 12 months.
No mentions of costs or budgets. Agentic AI seems to be taking off as businesses get their data AI-ready. “65% plan to increase AI spending over the next 12 months.”
The last chart shows how tough it is to be small in this rapidly expanding AI world. Small company = small AI staff, small budgets = small investments = small productivity gains = smaller market share.
Back to GS:
AI’s labor market impact has become increasingly visible in recent months. We continue to observe employment drags in specific industries where AI-use cases have been established—such as marketing, graphic design, customer service, and some tech occupations—and gains in construction jobs exposed to data centers (up 315k since 2022 relative to broader construction employment trends).
Recently, the three-month averages in nonmanufacturing payrolls and unemployment data show a statistically significant relationship with AI adoption at the subsector level, with higher adoption associated with weaker job growth and higher unemployment.
Corporate layoffs attributed to AI affected 4.0k employees in September, bringing the total since May 2023 to 207k.
We continue to observe large impacts on labor productivity in the limited areas where generative AI has been deployed but only modest signs of economy-wide impacts.
Academic studies imply a 23% average uplift to productivity, while company anecdotes imply slightly larger efficiency gains of around 32%.
Industries with higher AI adoption rates are now showing a slight acceleration in productivity growth over the past year in official US data.
EARNINGS WATCH
We learned a lot last week:
- Anthropic is now making good money from AI. It’s operating income reached $1B in Q2, close to $4B annualized.
Obviously not to be straight lined but, as SemiAnalysis wrote: “to date, they have burned $9.7b, they are growing at a rapid pace, and are profitable on an operating income basis as of the June quarter! This is likely the most efficient scaling of a business in the history of capitalism.”
- We also learned that hyperscaler cloud revenues not only keep accelerating (+54% YoY) but that their revenue backlog is up 140% to $1.75 Trillion!
(Goldman Sachs)
- We also learned that the data center business is quite profitable. Alphabet told us that its Google Cloud operating margin was 36.2% in Q3 (AMZN AWS’s was 38.7%), close to its Google Services margin of 40.7%. Its cloud revenues grew +103% YoY and its revenue backlog (now at $561bn) jumped 256% YoY. Management said that about 55% of this backlog (i.e. $310B) will convert into revenue in the next 24 months. Cloud revenues were $30.7B in Q3. It added that “additional availability of compute is standing as a gating factor to even faster growth”. FYI, Google Search revenues were $63B in Q3.
- Investors also told us that, in a world where financing capabilities are getting more challenging, they currently care more about free cashflow than AI growth.
(Goldman Sachs)
- Most interestingly, however, was that Alphabet, Amazon, Microsoft and Intel suffered from slowing, even declining, non-AI businesses:
- AMZN North America segment sales increased 6% after +16% in Q2 and +12% in Q1 (adjusting for the Prime Day shift, Q1+Q2 was +14.4% YoY and Q2+Q3 was +11.1%). On a world-wide basis, AMZN’s Net Product Sales slowed from +12.6% in Q1+Q2 to +7.5% in Q2+Q3. Meanwhile, AWS grew 39%, fastest pace in nearly five years.
- GOOG Search revenue growth slowed from +19% in Q1 and +17% in Q2 to +12% in Q3. Its cloud revenues grew +103%.
- MSFT Windows OEM and Devices division revenues cratered 22% in Q3 after +5% in Q2. Azure grew 48%.
- INTC Client Computing and Physical AI Group revenues dropped 5% after +13% in Q2. Intel’s data center revenues jumped 78%.
In effect, hyperscalers are now experiencing a K-shaped growth with their non-AI businesses, their cash cow divisions, slowing rather meaningfully, reflective of what’s happening in the non-AI economy.
- With some 180 S&P 500 companies reporting last week, we also learned that non-AI related revenues, margins and profits are starting to feel the effects of slowing non-AI demand and cost push inflation.
Goldman tells us that “the pressure from rising input cost prices has been reflected both in management commentary and in analyst margin estimate revisions.”
The latest PMI survey said that corporate costs are also surging due to “material shortages, low stock levels, overloaded suppliers, labour shortages, rising back orders and shipping or import delays”. That’s a lot to deal with all at once.
Non-AI profit growth is slowing even with surging war-related commodity producer profits (Energy +143% and Materials +40%).
In contrast with the last few quarters, the outlook for hyperscaler capex spending has remained fairly stable this reporting season. Analysts continue to expect roughly $800 billion in capex in 2026. Estimates for 2027 spending now stand at $1.2 trillion, only slightly above the estimate prior to the hyperscaler earnings reports.
The lack of revisions is consistent with our expectation that hyperscaler capex spending will increase substantially in 2027 but that we are moving past the peaks of both capex growth and upside surprises to consensus estimates.
Anthropic may not secure investment-grade rating after IPO, Blackstone tells investors Private capital giant says AI group is expected to receive ‘high’ junk rating in shift from previous guidance
Blackstone is acting as an anchor investor on the deal, known as “Project North Star”, and is working with banks that stumped up financing on selling the debt on to other investors. (…)
While Blackstone had told its investment partners in recent months that it was expecting an investment-grade rating for Anthropic before year-end, it has recently signalled that a rating in the double-B range or above is more likely given the guidance it has received from major credit agencies, the people said.
Such a designation would limit the pool of capital that Anthropic could tap because many insurance and pension funds are restricted from taking large positions in junk-rated borrowers. As a result, the company might need to continue to rely on its suppliers with high credit ratings to support its debt-raising efforts. (…)
The Smartest People in AI Believe This Is the World’s Most Important Chart AI is starting to build better AI. This self-improvement is speeding up research and development—and freaking everyone out.
(…) It’s a phenomenon known as “recursive self-improvement,” or RSI, which roughly translates to AI building smarter AI. Anthropic refers to it as “closing the loop,” a cycle in which each improvement makes the machines better and faster at improving themselves.
AI companies say they are now seeing the early signs of recursive self-improvement, which is speeding up their research—and freaking everyone out. As this technology accelerates its own evolution, the risk is that we struggle to keep pace and keep it under control.
Our best data about this profound shift comes from Anthropic, which recently published a chart showing how much of its AI research and development is being done by AI. (…)
US military says it helped Gulf states export 1.5 billion barrels of oil
US military forces have helped Gulf countries export more than 1.5 billion barrels of oil since June, despite Iran’s efforts to control the critical waterway, according to US Central Command. (…)
It added that more than 95 percent of ships navigating the channel had sailed through safely in recent months despite Iranian threats, “clearly demonstrating that the United States and regional partners control the international waterway”. (…)
At least 16.5 million barrels of oil per day left the region in September, equal to the pre-war average, excluding Iran, according to market intelligence firm Kpler.
Analysts have attributed the rebound to a combination of offshore ship-to-ship transfers, US military escorts, and increased use of oil pipelines. (…)
(…) The climbing prices reflect anxiety about the possibility of renewed hostilities in the US war on Iran, dwindling oil stocks, and the increased costs of transporting supplies out of the region, according to market analysts. (…)
Freight costs have reached as much as $40 per barrel in recent weeks, about 10 times the usual rate, according to data from the London Stock Exchange Group, the surge driven by elevated insurance premiums in addition to logistical challenges.
October has been the deadliest month for attacks on commercial vessels since the start of the war, with 32 incidents logged by the United Kingdom Maritime Trade Operations (UKMTO) as of Friday. (…)
Pavel Molchanov, an investment strategist at Raymond James & Associates, expressed scepticism that supply has returned to the pre-conflict trend.
“The IEA’s most recent production data – for September – shows OPEC Gulf production at 15.8 million bpd, which equates to 75 percent of the pre-war baseline,” Molchanov told Al Jazeera, referring to the International Energy Agency’s latest energy outlook. (…) “but it’s still a considerable distance from 100 percent.” (…)
Observed inventories fell by 10 million barrels in September, taking the global stockpile to 509 million barrels below its pre-war level, according to the IEA.
“The market has moved to a chronic squeeze, driven by concrete supply risks: Hormuz transit, shipping and depleted inventories,” Figuerola-Ferretti said. (…)
Goldman Sachs Group Inc. raised oil-price forecasts for the year-end and 2027 on elevated risks to supplies, joining other banks in hiking their outlooks after months of war in the Middle East.
Brent is now seen at $105 a barrel in December, up from a previous estimate of $90, and will average $88 next year, 10% more than seen before, analysts including Daan Struyven and Yulia Zhestkova Grigsby said in a note, which also increased forecasts for US oil marker West Texas Intermediate.
“Substantial downside supply risks are likely to sustain a large, near-term risk premium,” the analysts wrote. The markets are assigning “increasing weight to escalation after the Nov. 3 US midterms and subsequent potential weakness in Mideast supply,” they added in the Nov. 1 report. (…)
In recent weeks, Morgan Stanley raised its forecast for Dated Brent — a key physical price — to $105 a barrel for the fourth quarter, while ANZ Group Holdings Ltd. hiked its year-end outlook for Brent futures to the same figure.
EDGE and ODDS
From Callum Thomas:
- seasonality is about to flip maximally positive (though see the next chart for a caveat on this).
Source: @MikeZaccardi
- the monthly seasonality stats are revealing for a couple of reasons. First, November is the best month of the year from an average return standpoint and second-equal on % positive (historical 69% probability the market ends November up). Second, notice the best vs worst Novembers in history: +11% vs -11% (and then also contemplate that 31% of Novembers were negative). There is a seasonal edge for November (and December), but anything can happen to throw that off, and statistically there is almost a third chance that the market closes down. Be open to upside, but don’t rely on this as your only guide.
Source: Topdown Charts Pro
Meanwhile, the key dynamic in the post-ChatGPT era has been to bet on tech-led growth, which means buying tech stocks relative to bonds. The Nasdaq’s advance over Treasuries, proxied in the chart by the most popular exchange-traded funds tracking the sectors, has been remarkably consistent. September’s bond selloff took this trade beyond its long-term trend a month ago. After a few weeks of brutal correction, that trend remains intact:
Goldman Sachs: Previewing the Midterm Elections
(…) prediction markets imply high odds Democrats will win the House majority with a wider margin of control (around 236 seats) than either party has had in several years.
The Senate majority is more uncertain. (…) However, the Senate outcome is less important if Democrats have won the House; Senate passage generally requires bipartisan support, and President Trump could still veto any legislation Congress sends him.
Fiscal deadlines would be the main focus for legislative action under a Democratic majority. Spending authority will need to be extended and the debt limit will need to be raised, which could force compromise. The war in Iran, tariff rollback, the reversal of spending cuts, and the expansion of health benefits are likely to be at the top of the Democratic agenda around these deadlines. We don’t expect Congress to address these deadlines before year’s end, though Republicans might try.
We do not expect major fiscal policy changes over the next two years (…).
Democrats have indicated they would seek to reverse some of President Trump’s tariffs, but we expect this to have a limited effect since the reversals would likely be focused on a few countries (e.g., Canada) or products. We expect President Trump to announce new tariff increases after the election that would more than offset those reductions, likely resulting in a rise in the US effective tariff rate of a little under 2pp.
Authers:
Democrats’ chances of a Blue Wave are 70% — which is still less than the 80% shot Flavio Bolsonaro appeared to have against Lula in Brazil a week ago, but it still makes sense to call this a base case because they have a number of routes to their target.
For those keeping score on election night, the Democrats need to win a net four Senate seats, and Polymarket bettors think they’ll get six.
In order of confidence, they are North Carolina (98%), Ohio (74%), Alaska (74%), Texas (71%), Maine (68%), and Kansas (52%). Dems look more likely to win Iowa (42%) or even Nebraska (where an independent is on 27%), than to lose their most vulnerable state, Michigan (a 29% shot for the Republicans). You can entertain yourself by checking the latest odds on Polymarket’s website here. (…)
How much does all this matter to markets? The president is in place for another two years, and there is no chance of putting together a majority that could override his veto. Trump’s big tax cut for his second term has already happened, and he would veto any attempt to push through a hike.
A big Democratic win, coupled with victory for the billionaire tax proposal in California, would suggest the party might just win in 2028 with a candidate committed to raising taxes. That would be market-negative (to put it mildly); but Democrats would still need to choose a candidate and win the election (both difficult), so in the near-term this should do little more than affect the vibe. (…)
Congress cannot shift the market calculus significantly until the 2028 presidential election. A blue wave this week would change little in the short term, even if it shifted the odds and therefore the market prices for the longer term. (…)
If the GOP somehow held the House, it would be a massive vote of confidence in Trumpism and empower a radical end to his time in office. This hasn’t been priced in at all.
Barclays suggests that another tax cut would then be likely — and that would likely intensify both the rush to buy growth stocks and the upward pressure on inflation and interest rates. It would also be bullish for the dollar — the DXY dollar index has rallied after Trump victories over the last decade.
The likelihood is that a strong night for Democrats causes a mild freak-out over events that cannot happen for another two years. The slim Lula-level possibility of a last Trump hurrah would have much more impact.
AI impact:
Goldman thinks that
While data centers have emerged as a theme in some campaigns this year, we expect the focus to fade somewhat after the election. The issue ranks low among voter priorities, and state governors are likely to shift attention once election season is over. Congress might limit data center tax incentives, but a national moratorium looks unlikely in any election outcome. Instead, we expect the congressional focus to shift to AI regulation, where a compromise that codifies basic federal oversight and self-testing is possible but restrictive regulation looks unlikely.
But SemiAnalysis warns (my emphasis):
The 2027 datacenter boom is already being built and next week’s midterms won’t stop it now. (…) The election will change the rules around them, but much of the 2027 pipeline already has a substantial head start.
The fight is over 2028. Our model has 62.2 GW of US net additions that year, most of which hasn’t started construction. Most of it sits in states electing a governor on Nov. 3. (…)
That leaves a 2028 pipeline still working through approvals, power projects and construction schedules as the rules change. (…) What is changing is the reach of the restrictions given the growing anti-datacenter backlash.
Three things have worsened since.
- Texas, the largest US market, froze Texas Commission on Environmental Quality (TCEQ) permits on Sept. 21 on top of the interconnection-queue audit begun Aug. 3, bringing onsite power into the hold.
- Oregon expanded its earlier state-land freeze to air and water-discharge permits on Oct. 9, through July 1, 2027.
- And midterm candidates have campaigned on more restrictive positions: nine of the 20 nominees want a pause on new construction and the other 11 want conditions, and the split is about even by party. Both parties want developers to accept stricter conditions. The main difference is whether a developer can meet the new terms and keep building, or has to wait for the state to reopen approvals.
The biggest risk to our buildout forecast right now is a prolonged permitting freeze at the TCEQ. Texas alone adds close to 30% of US GW in 2028. A few months of lost time in that market can move the national number. (…) The question now is how long the pause lasts.
Gov. Greg Abbott (R), who is running for re-election in Texas, could reopen permitting once the Electric Reliability Council of Texas (ERCOT), the state’s grid operator, finishes its audit and the Texas Water Development Board completes its water-use review. ERCOT is targeting Dec. 10 to file its report, which does not by itself reopen permitting. A restart around then would leave developers dealing with a relatively short interruption, which some projects can absorb within their schedules. The audit still has to lead to a decision to resume issuance, and each permit still needs approval.
By contrast, Gina Hinojosa (D), his challenger, presents a more bearish approach: She would impose a moratorium until legislation is passed that gives communities a voice and protects household utility bills, as well as ends the tax incentives for datacenters. (…)
This is the risk posed by the pro-moratorium candidates: a project cannot carry out work if it still requires an approval which the state has ceased to issue. If the candidate proposing a datacenter pause won in every state where one is running, those states would together contain roughly 30 GW of planned capacity scheduled for 2028. How much actually slips depends on the permits each project still needs and how long the pause lasts.
We are less concerned with the candidates who support conditions only. Most of their demands change the price and terms of development. A developer can fund the grid upgrade, change the cooling design or negotiate the community agreement and then move forward. That can be expensive and time-consuming, but it gives projects a route to approval.
According to race ratings and available polling as of Oct. 30, from multiple sources (Cook Political Report, Multistate, and The New York Times), the conditions-only candidate is favored in Texas, Oklahoma, Pennsylvania, and Iowa, and Georgia is a toss-up.
(…) hyperscalers can afford to pay the higher costs of new rules and can more easily bring or pay for their own power.
However, that doesn’t necessarily mean they will be able to deliver capacity on time. These projects still require their own permits and equipment. And having the money to pay for power transmission doesn’t make it available immediately. (…)